Sensitivity analyses among subgroups encompassing participants without history of coronary heart disease (n = 1176), participants aged ≥40 years, the starting age of breast cancer screening in South Korea (n = 1140), and participants stratified by FRS risk
| Subgroups . | Sensitivity . | Specificity . | AUROC . | Accuracy (%) . |
|---|---|---|---|---|
| Participants without history of coronary heart disease (n = 1176) | ||||
| 0.763 (0.664–0.829) | 0.644 (0.607–0.707) | 0.758 (0.740–0.777) | 70.5 (68.5–72.1) | |
| Participants aged ≥40 years, the starting age of breast cancer screening in South Korea (n = 1140) | ||||
| 0.749 (0.658–0.822) | 0.674 (0.631–0.728) | 0.764 (0.743–0.785) | 71.1 (69.3–72.7) | |
| Stratified by FRS risk | ||||
| Participants at low risk: FRS <10% (n = 984) | ||||
| 0.640 (0.584–0.767) | 0.695 (0.591–0.737) | 0.717 (0.707–0.730) | 66.5 (64.7–69.4) | |
| Participants at higher risk: FRS ≥10% (n = 128) | ||||
| 0.760 (0.620–0.800) | 0.730 (0.683–0.814) | 0.718 (0.623–0.774) | 73.1 (68.8–81.4) | |
| Subgroups . | Sensitivity . | Specificity . | AUROC . | Accuracy (%) . |
|---|---|---|---|---|
| Participants without history of coronary heart disease (n = 1176) | ||||
| 0.763 (0.664–0.829) | 0.644 (0.607–0.707) | 0.758 (0.740–0.777) | 70.5 (68.5–72.1) | |
| Participants aged ≥40 years, the starting age of breast cancer screening in South Korea (n = 1140) | ||||
| 0.749 (0.658–0.822) | 0.674 (0.631–0.728) | 0.764 (0.743–0.785) | 71.1 (69.3–72.7) | |
| Stratified by FRS risk | ||||
| Participants at low risk: FRS <10% (n = 984) | ||||
| 0.640 (0.584–0.767) | 0.695 (0.591–0.737) | 0.717 (0.707–0.730) | 66.5 (64.7–69.4) | |
| Participants at higher risk: FRS ≥10% (n = 128) | ||||
| 0.760 (0.620–0.800) | 0.730 (0.683–0.814) | 0.718 (0.623–0.774) | 73.1 (68.8–81.4) | |
Abbreviations: AUROC, under the receiver operating characteristics; CAC, coronary artery calcium; FRS, Framingham risk score.
Performance was assessed using a prediction model based on a deep learning system employing a contrastive learning strategy, which utilized both positive and negative samples.
Sensitivity analyses among subgroups encompassing participants without history of coronary heart disease (n = 1176), participants aged ≥40 years, the starting age of breast cancer screening in South Korea (n = 1140), and participants stratified by FRS risk
| Subgroups . | Sensitivity . | Specificity . | AUROC . | Accuracy (%) . |
|---|---|---|---|---|
| Participants without history of coronary heart disease (n = 1176) | ||||
| 0.763 (0.664–0.829) | 0.644 (0.607–0.707) | 0.758 (0.740–0.777) | 70.5 (68.5–72.1) | |
| Participants aged ≥40 years, the starting age of breast cancer screening in South Korea (n = 1140) | ||||
| 0.749 (0.658–0.822) | 0.674 (0.631–0.728) | 0.764 (0.743–0.785) | 71.1 (69.3–72.7) | |
| Stratified by FRS risk | ||||
| Participants at low risk: FRS <10% (n = 984) | ||||
| 0.640 (0.584–0.767) | 0.695 (0.591–0.737) | 0.717 (0.707–0.730) | 66.5 (64.7–69.4) | |
| Participants at higher risk: FRS ≥10% (n = 128) | ||||
| 0.760 (0.620–0.800) | 0.730 (0.683–0.814) | 0.718 (0.623–0.774) | 73.1 (68.8–81.4) | |
| Subgroups . | Sensitivity . | Specificity . | AUROC . | Accuracy (%) . |
|---|---|---|---|---|
| Participants without history of coronary heart disease (n = 1176) | ||||
| 0.763 (0.664–0.829) | 0.644 (0.607–0.707) | 0.758 (0.740–0.777) | 70.5 (68.5–72.1) | |
| Participants aged ≥40 years, the starting age of breast cancer screening in South Korea (n = 1140) | ||||
| 0.749 (0.658–0.822) | 0.674 (0.631–0.728) | 0.764 (0.743–0.785) | 71.1 (69.3–72.7) | |
| Stratified by FRS risk | ||||
| Participants at low risk: FRS <10% (n = 984) | ||||
| 0.640 (0.584–0.767) | 0.695 (0.591–0.737) | 0.717 (0.707–0.730) | 66.5 (64.7–69.4) | |
| Participants at higher risk: FRS ≥10% (n = 128) | ||||
| 0.760 (0.620–0.800) | 0.730 (0.683–0.814) | 0.718 (0.623–0.774) | 73.1 (68.8–81.4) | |
Abbreviations: AUROC, under the receiver operating characteristics; CAC, coronary artery calcium; FRS, Framingham risk score.
Performance was assessed using a prediction model based on a deep learning system employing a contrastive learning strategy, which utilized both positive and negative samples.
This PDF is available to Subscribers Only
View Article Abstract & Purchase OptionsFor full access to this pdf, sign in to an existing account, or purchase an annual subscription.
